Named entity recognition

Problem

Customer-feedback and survey verbatims at Corebridge needed structured names, locations, and organizations before segmentation and CX models could use them reliably.

Constraints

• In-house evaluation set, not a public benchmark
• Feature layer for other models — latency and precision both mattered
• Domain language differs from generic spaCy defaults

What shipped

A custom spaCy NER model extracting person, location, and organization entities from feedback text, wired as a feature-engineering step into downstream segmentation and CX pipelines.

Outcome

95% precision and 90% recall on the internal eval set — good enough to trust as an upstream feature without drowning analysts in false entities.